Google's strides in computer vision leads to Google Lens feature

At Google I/O, CEO Sundar Pichai said that computer vision has reached an "inflection point," with error rates lower than a human's.
Written by Stephanie Condon, Senior Writer

At the Google I/O conference on Wednesday, Google unveiled a new feature bringing computer vision capabilities to various products, starting with Google Assistant and Google Photos later this year.

Called Google Lens, the feature helps you "understand what you're looking at" and take actions based on that information, CEO Sundar Pichai explained. For instance, you could point your phone at a flower and learn what kind of flower it is. You could point your phone at a restaurant and get contextual information such as its hours of operation. In another example, Pichai said, you could take a picture of a router and rather than typing in the password, "we can automatically do the hard work for you."

As Google strives to integrate artificial intelligence into all of its products, Google Lens illustrates how far computer vision in particular has come. In fact, the vision error rate of computer vision algorithms is now better than the human error rate.

Pichai called this "clearly at an inflection point with vision."

"The fact computers can understand images and videos has profound implications for our core mission" of organizing the world's information, he added.


Along with computer vision, Pichai gave an update on Google's speech-recognition capabilities, which now has a word error rate of 4.9 percent. He highlighted how advancements in speech recognition are influencing product design: Google initially planned to ship Google Home with eight microphones, but after applying deep learning, they were able to ship it with just two microphones and achieve the same quality. Thanks to deep learning, Google was able to recently announce support for multiple users in Google Home.

With its focus now squarely on artificial intelligence, Google also announced Wednesday that it's launched Google.ai, a collection of all the efforts and teams from across the company related to AI. It focuses on three areas: State of the art research, tools and infrastructure, and applied AI.

Google is "making impressive progress in applying machine learning and applying it across all products," Pichai said.

In the meantime, the company continues to build its impressive reach across the globe. Several of Google's major products and platforms now have more than 1 billion users, Pichai said, including YouTube, Maps, Chrome, Gmail, Search, and Play. Meanwhile, Google Drive, launched five years ago, now has 800 million active users. Photos, launched two years ago, has more than 500 million active users.

Editorial standards